On extended energy-loss fine structure data analysis for obtaining reliable structural parameters

We examined extended energy-loss fine structure (EXELFS) data analysis for obtaining reliable values of structural parameters, and its practical guide lines are presented. We found that the ratio method is unstable and gives r eliable results only when the calculated radial distribution function shows good peak separation, whereas the curve-fitting is rather robust for the q uantitative analysis. We introduced a simple optimization method to avoid t he instability associated with the ratio method so that two experiments on the same area at different temperatures should give time same coordination number. The coordination number and Debye-Waller factor derived from the ar ea having different thickness resulted in the same values within the experi mental accuracy up to the thickness of 0.4 lambda (lambda: plasmon mean fre e path), while the inter-atomic distance was linearly decreased with sample thickness also up to that thickness. This suggests that EXELFS experiments must be conducted for sample thickness <0.4 lambda.